The observations were carried out with the 2.7 m telescope of the McDonald Observatory, University of Texas, during two observing runs, March 1-9, 1994 and April 28 - May 2, 1994, using the 2d-coudé cross dispersion echelle spectrometer (Tull et al. 1995).
At the time of the observations the new 2k 
2k chip, which
covers most of the visual spectrum of the observed star in one
exposure, was not yet available at the spectrometer. The CCD-chip
used has 800 864 pixels and covers approximately 18 echelle
orders of length 25 - 35 Å. Settings of the detector for the
observations were selected to enable reliable analysis of the
spectral lines from the elements oxygen, sodium, magnesium and
aluminium, silicon, calcium, titanium, as well as chromium, iron,
nickel, yttrium and europium. Many sources were utilized to make the
selection of lines optimized for the current project, the most
important being Edvardsson et al. (1993a), Morell (1994),
#M&Mäckle et al. (1975), Ruland et al. (1981).
The Arcturus Atlas by Griffin (1968) and the Solar Spectrum Atlas by
Delbouille et al. (1973),
were carefully inspected, in order to avoid blends and define
suitable continuum regions. Altogether 3 different settings of the CCD
in the focal plane of the spectrometer camera were used.
The nominal spectral resolution (slit width 0.25 
m) of 
100000, was verified by observations of a thorium-argon lamp and
by measuring telluric lines. This relatively high resolution was
judged to be important in view of the relatively crowded spectra, in
particular for the cooler stars.
The reductions were carried out with standard IRAF
packages
for reduction and extraction of echelle spectra. The procedure
adopted was: all raw frames were first corrected for read-out noise
and bias, the flat fields added together, all object frames were
trimmed, the new flat field frames were normalized, the stellar
frames were divided by the normalized flat field, corrected for
scattered light and the one-dimensional stellar spectra
extracted. Finally, the spectrum of the comparison lamp, a
thorium-argon lamp, was reduced, extracted and the lines in it
identified. The comparison spectra were used to carry out the
wavelength calibration of the stellar spectra. A Legendre polynomial
was fitted to each stellar spectrum to define the continuum, using
the CONTINUUM task in IRAF.
The results of this procedure for defining the continuum were inspected visually and judged to give good continua for all the lines used in the abundance analysis; however, for the strong CaI line at 6162 Å, used for checking surface gravities, this definition of the continuum was not satisfactory for all stars. Instead, a blaze function was constructed for each night by fitting a Legendre polynomial to the spectrum of a B star observed during the night. The stellar spectra were then divided by this blaze function. Then a low order Legendre polynomial was fitted to the stellar spectrum to rectify it. The change in continuum was negligible for half of the stars, as compared with the result of using CONTINUUM; this also includes half of the K dwarf stars.
Our spectra are generally of high quality, reaching S/N of 200
in most cases. Some problems with fringes in the red were
encountered. The measurements of spectral line equivalent widths were
done using the SPLOT task in IRAF. Karin Eriksson is gratefully
thanked for carrying out almost half of the measurements. Care was
exercised to guarantee that the measurements agreed well between the
two operators.
The solar observations were performed by illuminating the slit of the
spectrograph by sky light each afternoon. The solar spectra have
2 times higher S/N ratios than the stellar spectra. The
reductions and measurements of the solar spectra were made in the same
way as for the stellar spectra.
The solar observations were used to determine "astrophysical'' 
-values. For lines, denoted by K in Table 2 (click here), shifted
by the difference in radial velocities between the stars and the Sun
outside our solar spectral recordings, but judged to be of interest
for the analyses, solar equivalent widths were instead obtained from
the solar spectrum atlas by Kurucz et al. (1984).